NANCHANG, July 29 (Xinhua) -- In east China's Jiangxi Province, a group of ceramic researchers has been restoring dragon vats, a type of vessel considered one of the most challenging large ceramic vessels in the country's ancient porcelain-making history.
Thanks to artificial intelligence (AI), tens of thousands of ancient ceramic shards unearthed in Jiangxi's Jingdezhen, well-known as China's "porcelain capital," have been pieced back together into stunning vessels, allowing visitors to behold the splendor of China's porcelain history.
The United Nations Educational, Scientific and Cultural Organization (UNESCO) added Jingdezhen Handicraft Porcelain Industry Sites to the World Heritage List on Saturday.
Dragon vats were porcelain pieces exclusively made for the imperial courts during the Ming Dynasty (1368-1644) and the Qing Dynasty (1644-1911) by the imperial kilns of ancient Jingdezhen. They were named for the cloud-dragon motifs painted on their walls. Firing such large porcelain vessels required extremely high standards, with only 5 out of 100 finished pieces passing inspection.
According to the regulations of that time, all rejected defective pieces had to be completely smashed and buried on-site, with entry into civilian circulation strictly forbidden. In the Ming Dynasty strata beneath the imperial kilns in Jingdezhen, over 15,000 shards left from the smashed vats have been discovered.
In a restoration studio of the Jingdezhen Imperial Kiln Institute, researchers have been working to restore these shards of varying sizes into vats. Since the discovery of these fragments, only three vats have been restored. One of the restored vats stands roughly 75 centimeters and is assembled from over 100 shards.
According to Weng Yanjun, director of the institute, ceramic restoration is a lengthy process that encompasses multiple stages, such as shard matching, cleaning and filling. Identifying the correct shards from a massive collection for matching is the most difficult aspect of the process.
In the traditional method, manual visual comparison was inefficient. Repeated handling would also easily wear down the fragile cross-sections of the shards, often causing irreversible secondary damage.
Through AI-driven matching and shard-based database archiving, the restoration process has been accelerated.
"What AI is doing now is helping us with the matching process, which is the most difficult step. Next, we want to expand its dataset while extending the three-dimensional spatial mapping of the dragon vats, ultimately assembling more complete ones," said Weng.
The data for AI intelligent matching comes from the world's first ancient ceramics database established in Jingdezhen. In June 2022, the Jingdezhen Imperial Kiln Institute, in collaboration with Tsinghua University, Peking University, the Palace Museum, the Shanghai Institute of Ceramics under the Chinese Academy of Sciences, and other institutions, established this database.
Researchers selected 3,000 sets comprising 12,000 specimens from the 20 million shards excavated at Jingdezhen. Each ancient specimen undergoes a series of procedures, including cleaning, cutting, and scanning electron microscopy analysis, resulting in four morphological states: shard, cross-section, thin section, and powder. Multiple precision instruments are then used to analyze microscopic morphology, body and glaze composition, and molecular structure.
The final database aggregates millions of data entries, covering core information such as archaeological data, shape, decoration, body, glaze, color, firing technique, and marks of the shards, as well as extended information including comparable collections in major museums worldwide, relevant archaeological discoveries at home and abroad, and modern academic research findings.
Weng noted that the team spent nearly 40 years manually matching and restoring three dragon vats. Using AI, they have already restored one and a half vats in less than two months.
"Manual restoration isn't just time-consuming. The human brain has reached its limit; it's difficult for humans alone to match them. AI technology has broken this deadlock," said Weng. ■
